top of page
Writer's pictureMatthew Lerner

Using Meta's AI to find Product/Market Fit

Meta offers the world’s most powerful AI for finding product/market fit, but 99% of us don’t know how to use it.


Claude and ChatGPT can tighten up your copy— but they don’t know your customers any better than you do. But Meta sure does.


Meta’s targeting engine, Lattice AI, has thousands of data points on 3.6 billion users, across 10 years of training data, including full-funnel conversions.


It can figure out who is likely to buy your product and which messages will resonate, and you can use that information to find answers to your company’s hardest questions, like:

  • Who are my ideal customers?

  • Which benefits matter most to them?

  • What’s the best way to explain my product’s value?


But that’s just the beginning. Of course you can use those answers to improve your ads. But you can also apply them to your landing pages, your onboarding flows, your cold emails, and even your product itself. (Plus, Meta can help you find other channels so you don’t have to depend on, well, Meta 😏)


My friend Hannah Parvaz, founder of Aperture, showed me this process, and here’s how it works:


1. Synthesize your customer insights into tight propositions

  • Gather all your customer insights from your Jobs to be Done interviews: Struggles, desired outcomes, alternative solutions, and pervasive anxieties about the category – anything that would appeal to a large swath of customers.

  • Narrow the list to 6-8 distinct themes to test, and develop messages around each one. For example, if you were selling healthy snacks, try some desired outcomes you heard from customers, like losing weight or having more energy. Or maybe attack competitors for their unhealthy ingredients. Create and test a message for each theme, and remember: Good messages are specific and use customers’ language.

  • Design one ad per theme, keeping the visuals similar, varying only the message. Five-second-test each ad variant before you run them to ensure the messages are readable and comprehensible.


2. Let the algo do its magic

  • Test 3-4 themes per week without a control group. Leave targeting wide open and use audience expansion. Hannah recommends running campaigns for 7 days or until you reach statistical significance, whichever comes later. Budget enough for 50 conversions per variant. (If your budget is limited, choose an optimization event higher up the funnel, but this carries some risk.)

  • Note: This AI-driven algorithm wants to drive efficient conversions, so it will search all possible users and identify the ones who are most attracted to each of your propositions.

  • Use the first round of experiments to narrow down to 3 effective angles. Then develop and test new concepts (“challengers”) against each. Iterate and refine until you have some ads working really well. Normally, one theme will emerge as the winner for driving full-funnel conversions.


3. Extract and apply the learnings:

  • Once you have a theme working, use your winning message to improve the rest of the funnel. Rework landing pages, app store listings, onboarding flows, welcome emails, cold outreach messages, etc. to echo your winning message.

  • Make sure your product delivers on that promise, and demonstrates that fact from the moment people start using it.

  • Bonus: Interview these new customers, again using the Jobs to be Done technique to discover other channels, so you won’t have to depend on paid ads. (I guarantee, Meta won’t be your prospects’ only information source. Find out what else they read, who they follow, what they Google, how they heard about your competitors.)


Meta’s AI-powered algorithm can do a lot of the heavy lifting of finding product-market fit: It can find customers, discover and validate your proposition. But it cannot apply those insights to the rest of your business – that’s down to you.


Just remember: Acquiring customers from Meta Ads? Pricey. Acquiring customer insights from Meta Ads? Priceless.

102 views0 comments

Recent Posts

See All

The 6 dysfunctions of OKRs

Unpopular opinion: OKRs don't cause problems, they surface them. If you have a clear strategy, and everyone understands the company's...

Commentaires


bottom of page